1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 2012, 2015 Chelsio Communications, Inc. 5 * All rights reserved. 6 * Written by: Navdeep Parhar <np@FreeBSD.org> 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * $FreeBSD$ 30 * 31 */ 32 33 #ifndef __T4_TOM_H__ 34 #define __T4_TOM_H__ 35 #include <sys/vmem.h> 36 #include "common/t4_hw.h" 37 #include "tom/t4_tls.h" 38 39 #define LISTEN_HASH_SIZE 32 40 41 /* 42 * Min receive window. We want it to be large enough to accommodate receive 43 * coalescing, handle jumbo frames, and not trigger sender SWS avoidance. 44 */ 45 #define MIN_RCV_WND (24 * 1024U) 46 47 /* 48 * Max receive window supported by HW in bytes. Only a small part of it can 49 * be set through option0, the rest needs to be set through RX_DATA_ACK. 50 */ 51 #define MAX_RCV_WND ((1U << 27) - 1) 52 53 #define DDP_RSVD_WIN (16 * 1024U) 54 #define SB_DDP_INDICATE SB_IN_TOE /* soreceive must respond to indicate */ 55 56 #define USE_DDP_RX_FLOW_CONTROL 57 58 #define PPOD_SZ(n) ((n) * sizeof(struct pagepod)) 59 #define PPOD_SIZE (PPOD_SZ(1)) 60 61 /* TOE PCB flags */ 62 enum { 63 TPF_ATTACHED = (1 << 0), /* a tcpcb refers to this toepcb */ 64 TPF_FLOWC_WR_SENT = (1 << 1), /* firmware flow context WR sent */ 65 TPF_TX_DATA_SENT = (1 << 2), /* some data sent */ 66 TPF_TX_SUSPENDED = (1 << 3), /* tx suspended for lack of resources */ 67 TPF_SEND_FIN = (1 << 4), /* send FIN after all pending data */ 68 TPF_FIN_SENT = (1 << 5), /* FIN has been sent */ 69 TPF_ABORT_SHUTDOWN = (1 << 6), /* connection abort is in progress */ 70 TPF_CPL_PENDING = (1 << 7), /* haven't received the last CPL */ 71 TPF_SYNQE = (1 << 8), /* synq_entry, not really a toepcb */ 72 TPF_SYNQE_EXPANDED = (1 << 9), /* toepcb ready, tid context updated */ 73 TPF_FORCE_CREDITS = (1 << 10), /* always send credits */ 74 }; 75 76 enum { 77 DDP_OK = (1 << 0), /* OK to turn on DDP */ 78 DDP_SC_REQ = (1 << 1), /* state change (on/off) requested */ 79 DDP_ON = (1 << 2), /* DDP is turned on */ 80 DDP_BUF0_ACTIVE = (1 << 3), /* buffer 0 in use (not invalidated) */ 81 DDP_BUF1_ACTIVE = (1 << 4), /* buffer 1 in use (not invalidated) */ 82 DDP_TASK_ACTIVE = (1 << 5), /* requeue task is queued / running */ 83 DDP_DEAD = (1 << 6), /* toepcb is shutting down */ 84 }; 85 86 struct sockopt; 87 struct offload_settings; 88 89 struct ofld_tx_sdesc { 90 uint32_t plen; /* payload length */ 91 uint8_t tx_credits; /* firmware tx credits (unit is 16B) */ 92 void *iv_buffer; /* optional buffer holding IVs for TLS */ 93 }; 94 95 struct ppod_region { 96 u_int pr_start; 97 u_int pr_len; 98 u_int pr_page_shift[4]; 99 uint32_t pr_tag_mask; /* hardware tagmask for this region. */ 100 uint32_t pr_invalid_bit; /* OR with this to invalidate tag. */ 101 uint32_t pr_alias_mask; /* AND with tag to get alias bits. */ 102 u_int pr_alias_shift; /* shift this much for first alias bit. */ 103 vmem_t *pr_arena; 104 }; 105 106 struct ppod_reservation { 107 struct ppod_region *prsv_pr; 108 uint32_t prsv_tag; /* Full tag: pgsz, alias, tag, color */ 109 u_int prsv_nppods; 110 }; 111 112 struct pageset { 113 TAILQ_ENTRY(pageset) link; 114 vm_page_t *pages; 115 int npages; 116 int flags; 117 int offset; /* offset in first page */ 118 int len; 119 struct ppod_reservation prsv; 120 struct vmspace *vm; 121 vm_offset_t start; 122 u_int vm_timestamp; 123 }; 124 125 TAILQ_HEAD(pagesetq, pageset); 126 127 #define PS_PPODS_WRITTEN 0x0001 /* Page pods written to the card. */ 128 129 struct ddp_buffer { 130 struct pageset *ps; 131 132 struct kaiocb *job; 133 int cancel_pending; 134 }; 135 136 struct ddp_pcb { 137 u_int flags; 138 struct ddp_buffer db[2]; 139 TAILQ_HEAD(, pageset) cached_pagesets; 140 TAILQ_HEAD(, kaiocb) aiojobq; 141 u_int waiting_count; 142 u_int active_count; 143 u_int cached_count; 144 int active_id; /* the currently active DDP buffer */ 145 struct task requeue_task; 146 struct kaiocb *queueing; 147 struct mtx lock; 148 }; 149 150 struct toepcb { 151 TAILQ_ENTRY(toepcb) link; /* toep_list */ 152 u_int flags; /* miscellaneous flags */ 153 int refcount; 154 struct tom_data *td; 155 struct inpcb *inp; /* backpointer to host stack's PCB */ 156 struct vnet *vnet; 157 struct vi_info *vi; /* virtual interface */ 158 struct sge_wrq *ofld_txq; 159 struct sge_ofld_rxq *ofld_rxq; 160 struct sge_wrq *ctrlq; 161 struct l2t_entry *l2te; /* L2 table entry used by this connection */ 162 struct clip_entry *ce; /* CLIP table entry used by this tid */ 163 int tid; /* Connection identifier */ 164 int tc_idx; /* traffic class that this tid is bound to */ 165 166 /* tx credit handling */ 167 u_int tx_total; /* total tx WR credits (in 16B units) */ 168 u_int tx_credits; /* tx WR credits (in 16B units) available */ 169 u_int tx_nocompl; /* tx WR credits since last compl request */ 170 u_int plen_nocompl; /* payload since last compl request */ 171 172 uint16_t opt0_rcv_bufsize; /* XXX: save full opt0/opt2 for later? */ 173 uint16_t mtu_idx; 174 uint16_t emss; 175 uint16_t tcp_opt; 176 177 u_int ulp_mode; /* ULP mode */ 178 void *ulpcb; 179 void *ulpcb2; 180 struct mbufq ulp_pduq; /* PDUs waiting to be sent out. */ 181 struct mbufq ulp_pdu_reclaimq; 182 183 struct ddp_pcb ddp; 184 struct tls_ofld_info tls; 185 186 TAILQ_HEAD(, kaiocb) aiotx_jobq; 187 struct task aiotx_task; 188 struct socket *aiotx_so; 189 190 /* Tx software descriptor */ 191 uint8_t txsd_total; 192 uint8_t txsd_pidx; 193 uint8_t txsd_cidx; 194 uint8_t txsd_avail; 195 struct ofld_tx_sdesc txsd[]; 196 }; 197 198 #define DDP_LOCK(toep) mtx_lock(&(toep)->ddp.lock) 199 #define DDP_UNLOCK(toep) mtx_unlock(&(toep)->ddp.lock) 200 #define DDP_ASSERT_LOCKED(toep) mtx_assert(&(toep)->ddp.lock, MA_OWNED) 201 202 struct flowc_tx_params { 203 uint32_t snd_nxt; 204 uint32_t rcv_nxt; 205 unsigned int snd_space; 206 unsigned int mss; 207 }; 208 209 /* 210 * Compressed state for embryonic connections for a listener. 211 */ 212 struct synq_entry { 213 struct listen_ctx *lctx; /* backpointer to listen ctx */ 214 struct mbuf *syn; 215 int flags; /* same as toepcb's tp_flags */ 216 volatile int ok_to_respond; 217 volatile u_int refcnt; 218 int tid; 219 uint32_t iss; 220 uint32_t irs; 221 uint32_t ts; 222 uint16_t txqid; 223 uint16_t rxqid; 224 uint16_t l2e_idx; 225 uint16_t ulp_mode; 226 uint16_t rcv_bufsize; 227 __be16 tcp_opt; /* from cpl_pass_establish */ 228 struct toepcb *toep; 229 }; 230 231 /* listen_ctx flags */ 232 #define LCTX_RPL_PENDING 1 /* waiting for a CPL_PASS_OPEN_RPL */ 233 234 struct listen_ctx { 235 LIST_ENTRY(listen_ctx) link; /* listen hash linkage */ 236 volatile int refcount; 237 int stid; 238 struct stid_region stid_region; 239 int flags; 240 struct inpcb *inp; /* listening socket's inp */ 241 struct vnet *vnet; 242 struct sge_wrq *ctrlq; 243 struct sge_ofld_rxq *ofld_rxq; 244 struct clip_entry *ce; 245 }; 246 247 /* tcb_histent flags */ 248 #define TE_RPL_PENDING 1 249 #define TE_ACTIVE 2 250 251 /* bits in one 8b tcb_histent sample. */ 252 #define TS_RTO (1 << 0) 253 #define TS_DUPACKS (1 << 1) 254 #define TS_FASTREXMT (1 << 2) 255 #define TS_SND_BACKLOGGED (1 << 3) 256 #define TS_CWND_LIMITED (1 << 4) 257 #define TS_ECN_ECE (1 << 5) 258 #define TS_ECN_CWR (1 << 6) 259 #define TS_RESERVED (1 << 7) /* Unused. */ 260 261 struct tcb_histent { 262 struct mtx te_lock; 263 struct callout te_callout; 264 uint64_t te_tcb[TCB_SIZE / sizeof(uint64_t)]; 265 struct adapter *te_adapter; 266 u_int te_flags; 267 u_int te_tid; 268 uint8_t te_pidx; 269 uint8_t te_sample[100]; 270 }; 271 272 struct tom_data { 273 struct toedev tod; 274 275 /* toepcb's associated with this TOE device */ 276 struct mtx toep_list_lock; 277 TAILQ_HEAD(, toepcb) toep_list; 278 279 struct mtx lctx_hash_lock; 280 LIST_HEAD(, listen_ctx) *listen_hash; 281 u_long listen_mask; 282 int lctx_count; /* # of lctx in the hash table */ 283 284 struct ppod_region pr; 285 286 struct rwlock tcb_history_lock __aligned(CACHE_LINE_SIZE); 287 struct tcb_histent **tcb_history; 288 int dupack_threshold; 289 290 /* WRs that will not be sent to the chip because L2 resolution failed */ 291 struct mtx unsent_wr_lock; 292 STAILQ_HEAD(, wrqe) unsent_wr_list; 293 struct task reclaim_wr_resources; 294 }; 295 296 static inline struct tom_data * 297 tod_td(struct toedev *tod) 298 { 299 300 return (__containerof(tod, struct tom_data, tod)); 301 } 302 303 static inline struct adapter * 304 td_adapter(struct tom_data *td) 305 { 306 307 return (td->tod.tod_softc); 308 } 309 310 static inline void 311 set_mbuf_ulp_submode(struct mbuf *m, uint8_t ulp_submode) 312 { 313 314 M_ASSERTPKTHDR(m); 315 m->m_pkthdr.PH_per.eight[0] = ulp_submode; 316 } 317 318 static inline uint8_t 319 mbuf_ulp_submode(struct mbuf *m) 320 { 321 322 M_ASSERTPKTHDR(m); 323 return (m->m_pkthdr.PH_per.eight[0]); 324 } 325 326 /* t4_tom.c */ 327 struct toepcb *alloc_toepcb(struct vi_info *, int, int, int); 328 struct toepcb *hold_toepcb(struct toepcb *); 329 void free_toepcb(struct toepcb *); 330 void offload_socket(struct socket *, struct toepcb *); 331 void undo_offload_socket(struct socket *); 332 void final_cpl_received(struct toepcb *); 333 void insert_tid(struct adapter *, int, void *, int); 334 void *lookup_tid(struct adapter *, int); 335 void update_tid(struct adapter *, int, void *); 336 void remove_tid(struct adapter *, int, int); 337 int find_best_mtu_idx(struct adapter *, struct in_conninfo *, 338 struct offload_settings *); 339 u_long select_rcv_wnd(struct socket *); 340 int select_rcv_wscale(void); 341 uint64_t calc_opt0(struct socket *, struct vi_info *, struct l2t_entry *, 342 int, int, int, int, struct offload_settings *); 343 uint64_t select_ntuple(struct vi_info *, struct l2t_entry *); 344 int select_ulp_mode(struct socket *, struct adapter *, 345 struct offload_settings *); 346 void set_ulp_mode(struct toepcb *, int); 347 int negative_advice(int); 348 int add_tid_to_history(struct adapter *, u_int); 349 350 /* t4_connect.c */ 351 void t4_init_connect_cpl_handlers(void); 352 void t4_uninit_connect_cpl_handlers(void); 353 int t4_connect(struct toedev *, struct socket *, struct rtentry *, 354 struct sockaddr *); 355 void act_open_failure_cleanup(struct adapter *, u_int, u_int); 356 357 /* t4_listen.c */ 358 void t4_init_listen_cpl_handlers(void); 359 void t4_uninit_listen_cpl_handlers(void); 360 int t4_listen_start(struct toedev *, struct tcpcb *); 361 int t4_listen_stop(struct toedev *, struct tcpcb *); 362 void t4_syncache_added(struct toedev *, void *); 363 void t4_syncache_removed(struct toedev *, void *); 364 int t4_syncache_respond(struct toedev *, void *, struct mbuf *); 365 int do_abort_req_synqe(struct sge_iq *, const struct rss_header *, 366 struct mbuf *); 367 int do_abort_rpl_synqe(struct sge_iq *, const struct rss_header *, 368 struct mbuf *); 369 void t4_offload_socket(struct toedev *, void *, struct socket *); 370 void synack_failure_cleanup(struct adapter *, int); 371 372 /* t4_cpl_io.c */ 373 void aiotx_init_toep(struct toepcb *); 374 int t4_aio_queue_aiotx(struct socket *, struct kaiocb *); 375 void t4_init_cpl_io_handlers(void); 376 void t4_uninit_cpl_io_handlers(void); 377 void send_abort_rpl(struct adapter *, struct sge_wrq *, int , int); 378 void send_flowc_wr(struct toepcb *, struct flowc_tx_params *); 379 void send_reset(struct adapter *, struct toepcb *, uint32_t); 380 int send_rx_credits(struct adapter *, struct toepcb *, int); 381 void send_rx_modulate(struct adapter *, struct toepcb *); 382 void make_established(struct toepcb *, uint32_t, uint32_t, uint16_t); 383 int t4_close_conn(struct adapter *, struct toepcb *); 384 void t4_rcvd(struct toedev *, struct tcpcb *); 385 void t4_rcvd_locked(struct toedev *, struct tcpcb *); 386 int t4_tod_output(struct toedev *, struct tcpcb *); 387 int t4_send_fin(struct toedev *, struct tcpcb *); 388 int t4_send_rst(struct toedev *, struct tcpcb *); 389 void t4_set_tcb_field(struct adapter *, struct sge_wrq *, struct toepcb *, 390 uint16_t, uint64_t, uint64_t, int, int); 391 void t4_push_frames(struct adapter *sc, struct toepcb *toep, int drop); 392 void t4_push_pdus(struct adapter *sc, struct toepcb *toep, int drop); 393 394 /* t4_ddp.c */ 395 int t4_init_ppod_region(struct ppod_region *, struct t4_range *, u_int, 396 const char *); 397 void t4_free_ppod_region(struct ppod_region *); 398 int t4_alloc_page_pods_for_ps(struct ppod_region *, struct pageset *); 399 int t4_alloc_page_pods_for_buf(struct ppod_region *, vm_offset_t, int, 400 struct ppod_reservation *); 401 int t4_write_page_pods_for_ps(struct adapter *, struct sge_wrq *, int, 402 struct pageset *); 403 int t4_write_page_pods_for_buf(struct adapter *, struct sge_wrq *, int tid, 404 struct ppod_reservation *, vm_offset_t, int); 405 void t4_free_page_pods(struct ppod_reservation *); 406 int t4_soreceive_ddp(struct socket *, struct sockaddr **, struct uio *, 407 struct mbuf **, struct mbuf **, int *); 408 int t4_aio_queue_ddp(struct socket *, struct kaiocb *); 409 void t4_ddp_mod_load(void); 410 void t4_ddp_mod_unload(void); 411 void ddp_assert_empty(struct toepcb *); 412 void ddp_init_toep(struct toepcb *); 413 void ddp_uninit_toep(struct toepcb *); 414 void ddp_queue_toep(struct toepcb *); 415 void release_ddp_resources(struct toepcb *toep); 416 void handle_ddp_close(struct toepcb *, struct tcpcb *, uint32_t); 417 void handle_ddp_indicate(struct toepcb *); 418 void insert_ddp_data(struct toepcb *, uint32_t); 419 const struct offload_settings *lookup_offload_policy(struct adapter *, int, 420 struct mbuf *, uint16_t, struct inpcb *); 421 422 /* t4_tls.c */ 423 bool can_tls_offload(struct adapter *); 424 int t4_ctloutput_tls(struct socket *, struct sockopt *); 425 void t4_push_tls_records(struct adapter *, struct toepcb *, int); 426 void t4_tls_mod_load(void); 427 void t4_tls_mod_unload(void); 428 void tls_establish(struct toepcb *); 429 void tls_init_toep(struct toepcb *); 430 int tls_rx_key(struct toepcb *); 431 void tls_stop_handshake_timer(struct toepcb *); 432 int tls_tx_key(struct toepcb *); 433 void tls_uninit_toep(struct toepcb *); 434 435 #endif 436